データ前処理

データの前処理に関するコードメモ

#データフレーム内の欠損値数の確認
df_train.isnull().sum()
 ages                  18
list_price             0
num_reviews            0
piece_count            0
review_difficulty      0
star_rating          405
country              255
dtype: int64

#null値を持つ行を削除
df = df.dropna(subset=['charges'])

#欠損値の補完（平均値or最頻値を補完）
df = df.fillna({'bmi': df['bmi'].mean()})

#カテゴリカル変数の確認
df_obj = df.select_dtypes(include='object')
gender	smoker	region
0	female	yes	southwest
3	male	no	northwest
9	female	no	northwest

df_uni = df_obj.nunique()
gender    2
smoker    2
region    4
dtype: int64

for uni in df_obj.columns:
    print(uni)
    print(df_obj[uni].unique())

gender
['female' 'male']
smoker
['yes' 'no']
region
['southwest' 'northwest' 'southeast' 'northeast']

#カテゴリカル変数をダミー変数へ
from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()
le.fit(df['gender'])
le.transform(df['gender'])

age	gender	bmi	children	smoker	region	charges
0	19	0	27.900000	0	yes	southwest	16884.92400
3	33	1	22.705000	0	no	northwest	21984.47061
9	60	0	30.716434	0	no	northwest	28923.13692

#One-Hot Encording (カテゴリカル変数をすべてダミー変数へ)
df = pd.get_dummies(df, drop_first=True)

One-Hot Encordingがどちゃくそ便利！